Some patients know within the first sentence; some never do; most sit in the middle — suspect but don't care once the call is resolved. How patients identify AI, and whether practices should disclose explicitly, is partly a design choice and increasingly a legal question. Several U.S. states are adding AI-disclosure requirements that take effect in 2026 and 2027. Here's what patients notice, what the emerging legal requirements look like, and how practices are handling disclosure successfully.
What Gives It Away
Modern voice AI is very good. The remaining tells, in rough order of frequency:
- Response speed and consistency: no "hold on, let me check" pauses that humans have
- Pronunciation of unusual names: a default pronunciation that's slightly off for names outside the common distribution
- Turn-taking timing: AI sometimes replies a half-beat sooner than a human would
- Unflappable tone: no traces of fatigue, impatience, or casual variability — perfectly consistent every call
- Scripted transitions: moving between topics smoothly but in ways that occasionally feel templated
In 2024 most patients could tell within 30 seconds. In 2026 most can't tell until they actively think about it.
The Emerging Legal Disclosure Landscape
California (AB 3 and related)
Effective 2026, California requires disclosure when AI agents interact with consumers in certain commercial contexts. Healthcare communication is generally within scope. Practices operating in California should include AI identification in the greeting.
New York (proposed), Texas, Illinois
Similar requirements proposed or enacted. Requirements vary — some mandate disclosure in the greeting; others require disclosure upon request; others require a written notice during the patient relationship. Legal counsel is recommended for multi-state practices.
Federal
FTC has issued guidance on AI disclosure in commercial contexts. No hard federal healthcare-specific rule yet, but the direction of travel is toward more disclosure, not less.
The practical, conservative path
Disclose in the greeting, regardless of state. Reduces legal risk, aligns with emerging norms, and most patients are fine with it.
How Practices Handle Disclosure
The explicit model
"You've reached Lakeside Dental. I'm Ava, our virtual coordinator — I can schedule, answer questions, or connect you with our team. How can I help you today?" Direct identification. Most patients move on to their question.
The soft-disclosure model
"Lakeside Dental front desk, how can I help?" — neutral, identifies as AI when asked. Growing less common as disclosure laws expand.
The hybrid model
"You've reached Lakeside Dental's front desk. How can I help you today?" → patient states request → "Great, I can help you with that. I'm Ava, Lakeside's virtual coordinator — let me get you scheduled." Disclosure a few turns in, after trust is established on content.
Patient response varies by model. Explicit tends to feel most honest; hybrid feels most natural. Soft risks legal exposure and patient trust if they later learn.
What Patients Actually Care About
Surveyed responses from practices using AI phone handling:
- "Was my issue resolved?" — yes: satisfaction high, AI vs. human secondary
- "Did I feel respected?" — tone and acknowledgment matter more than human-ness
- "Can I reach a human if I need one?" — knowing the escalation path matters more than talking to a human on this call
- "Was the experience efficient?" — AI typically wins on this metric
- "Do I feel weird about AI handling my health stuff?" — for most, no; for some, yes, and they appreciate the transfer option
How to Frame Disclosure So It Builds Trust
- Name the AI. "I'm Ava, our virtual coordinator." Using a name humanizes and makes the experience feel personal rather than corporate.
- State the value. "I'm here to help you book or connect you with the right person." The patient sees you're offering choice, not gating.
- Offer the off-ramp. "Just say 'speak to a team member' any time and I'll transfer you." Removes feeling of being trapped.
- Acknowledge the novelty. Some practices lean in: "We use AI to make sure every call gets answered. Real humans are here for complex stuff."
- Stay warm. AI can be warm without being chatty. Tone carries the weight.
The 3–5% Who Don't Want AI
A small fraction of patients will hang up or ask for a human immediately. That's fine. The AI transfers them. Lost patients are nearly zero if the transfer is quick. What loses patients is the AI refusing or stalling a transfer.
Updates to Patient-Facing Materials
Consider updating:
- Notice of Privacy Practices — reference automated phone handling
- Website "how we handle your information" section
- Welcome letters or new-patient emails — a sentence or two about calling in
- Front-desk signage — optional but sometimes appreciated by patients
FAQ
Do I have to disclose legally?
Depends on your state. Even where not required, disclosure is the conservative and trust-building choice.
What if patients ask "are you a real person?"
Answer honestly. "No, I'm a virtual coordinator — I can help with your question, or I can connect you with a real team member. Which would you prefer?" Any vendor whose AI deflects this question has a design problem.
Should we announce AI to existing patients?
A simple email or SMS the week before launch is best practice. "We've added Ava, our virtual coordinator, to answer calls 24/7. Reach our team anytime by saying 'team member.'" Brief, factual, no hype.
Does disclosure reduce bookings?
Slightly measurable drop for the small fraction who prefer humans. Net impact is clearly positive because the AI handles many more total calls than it loses to disclosure-aversion.
What about minors calling?
Same disclosure. Parents calling on behalf of minors appreciate clarity about who's handling their child's information.